Sometimes Even Google Misses Big

This morning a poor soul entered the folllowing keyword in Google: "how to build road on a hill doityourself."  He wanted to learn how to build a road on a hill and wanted to do it himself.  Maybe it's for his driveway, maybe he's doing it for a living.  His motivations are interesting, but not relevant.  Neither were the results he got from Google.

The most relevant result Google gave him, the #1 spot they offered was my blog post on vertical search engines called, Do It Yourself Vertical Search, where I criticize the value of a lot of the vertical search engines out there.  My blog title has the words "hill and road" in it, the post has "do it yourself" in it, so a post on vertical search is the most relevant result for somebody looking to build a road on a hill.  I don't think I helped this guy at all.

Now, I thought this was amusing, but it is by no means rare.  When you put an entry in Google or any search engine that's 7-10 words long you are almost always out of luck.  It is in these long keywords 3+ words that search engines have to prove their worth going forward.  Shorter keywords searches all give you ads anyway, no matter what side fo the screen they show up.

Other search engines couldn't help this guy build his hill better either.  Mahalo suggested "King of the Hill" episode guides, before reverting to Google, though at least they said they don't have a relevant result at the top.  Over time they will fix this, and the link above will become relevant.

Yahoo didn't do much better either.  They also found my post but also confused the poor guy doing the search by telling him this:  "We have included results for how to build road on a hill do it yourself. Do you want results just for how to build road on a hill doityourself?"  This might look like a one-up on Google's way of dealing with misspellings, but is all this confusion worth that extra click you save the guy?

Anyway, I thought this was amusing, but I guess there is a point and that when it comes to long keywords, generic search engines are failing, social engines (and products like Yahoo! answers) have a shot of filling the void, and if you really want to learn how to build a road on a hill, go to Home Depot.

Comments

[Adam Jusko]

Our human-powered search engine Bessed wouldn't have found the searcher a good answer, either, but your point about bad results for long tail searches is exactly the strategy we're going after. Google's fine on the most-searched phrases, but when you get into more detailed searches, it has a much harder time, and that's the area we're trying to focus on. Not that we'll forget the bigger searches, but we think the place humans can really make a difference in the search field is by finding the hard stuff that an algorithm can't winnow out.

[annplugged]

Baris, while I totally understand your amazement at such results, and feel the possible frustration of that DIY searcher, let me add some quick egg-headed remarks. In defense of search spiders, or sth like that. 1, all these blog analytics reflecting blogposts on longtail search results commented by the blog authors (which I quite often see in fact) will only reinforce search algorithms that you are the relevant hit for the above terms
2, your blog title is unique in a sense that your site gets extra algo points for having both hill and road (treated separately, not as 'Road Hill' property name). That only means, I think, that currently there are no sites that would work with 'hill' and 'road' in such emphatic position as your site (H1 metatag), which is a frequently crawled blog site to boost it. But it also means that your head title is fancy/whimsical, rather than descriptive
3, searchers who use full sentences are usually less experienced searchers. Now once Marissa (Mayer) said that based on the results they have collected so far, searchers learn to search better in one month. It is, in all probability, true, but I have always wanted to develop a search training series (I have helped lots of people with search tips) as I'm convinced that a one-month developmental arch suddenly drops. Users think 'Now I can search' and they stop discovering further tricks of the trade.

[Baris Karadogan]

Annplugged,
Thanks for this very insightful comment.